Your experiments have been most helpful. According to me, the fact that you have gotten this results, you have already succeeded: here is why...After seeing all these, I went back to my breadboard, I setup a 20KHz PWM using 555 timer, took my transformer and connected the 7v side to my mosfet bridge, I connected the 220v side to mains AC, without switching, I got 4v DC from mosfet bridge, I powered the breadboard and start switching, at first, no voltage increase, so, I started increasing the duty cycle starting from 10% till I got to 50%, and the voltage gradually increased from 4v to 5v, which is not what I was expecting.
I tried everything I could, I couldn't get pass 5v DC, i bumped up the frequency, still no difference, so I decided to change the output capacitor from 35v 100uF to 35v 1000uF, and I was able to get to 5.5v.
Then I changed it again to 16v 3300uF and I was able to get to 6.3v, so I added 3 of the 3300uF capacitors in parallel, but no difference. So, my best result was 6.3v
I don't know what is happening or what I am doing wrong.
In all the UPS I have dissected (>40 and >5 different brands all of them for 12v batteries), the transformer always have at least one extra input coil on the primary winding. The extra coil when fed to mains produces a voltage of about ~11vac while the one used by the inverter produces ~7vac. A relay is usually attached on the hot side to switch between the 2 input wires which are often from the same coil. The 11v coil when rectified should give >13.8vdc which in my thinking is sufficient assuming no boost regulation. With your 50% increase (4v to 6.3v) that would give you close to 20v offering you enough room to pwm for whatever current charging needs.